2 * Copyright (c) 1988 University of Utah.
3 * Copyright (c) 1990, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * the Systems Programming Group of the University of Utah Computer
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * from: Utah Hdr: vn.c 1.13 94/04/02
40 * from: @(#)vn.c 8.6 (Berkeley) 4/1/94
41 * $FreeBSD: src/sys/dev/vn/vn.c,v 1.105.2.4 2001/11/18 07:11:00 dillon Exp $
42 * $DragonFly: src/sys/dev/disk/vn/vn.c,v 1.35 2007/05/19 20:31:16 dillon Exp $
48 * Block/character interface to a vnode. Allows one to treat a file
49 * as a disk (e.g. build a filesystem in it, mount it, etc.).
51 * NOTE 1: There is a security issue involved with this driver.
52 * Once mounted all access to the contents of the "mapped" file via
53 * the special file is controlled by the permissions on the special
54 * file, the protection of the mapped file is ignored (effectively,
55 * by using root credentials in all transactions).
57 * NOTE 2: Doesn't interact with leases, should it?
60 #include <sys/param.h>
61 #include <sys/systm.h>
62 #include <sys/kernel.h>
64 #include <sys/nlookup.h>
66 #include <sys/malloc.h>
67 #include <sys/mount.h>
68 #include <sys/vnode.h>
69 #include <sys/fcntl.h>
71 #include <sys/diskslice.h>
74 #include <sys/module.h>
75 #include <sys/vnioctl.h>
78 #include <vm/vm_object.h>
79 #include <vm/vm_page.h>
80 #include <vm/vm_pager.h>
81 #include <vm/vm_pageout.h>
82 #include <vm/swap_pager.h>
83 #include <vm/vm_extern.h>
84 #include <vm/vm_zone.h>
86 static d_ioctl_t vnioctl;
87 static d_open_t vnopen;
88 static d_close_t vnclose;
89 static d_psize_t vnsize;
90 static d_strategy_t vnstrategy;
94 #define VN_BSIZE_BEST 8192
98 * D_DISK we want to look like a disk
99 * D_CANFREE We support BUF_CMD_FREEBLKS
102 static struct dev_ops vn_ops = {
103 { "vn", CDEV_MAJOR, D_DISK | D_CANFREE },
107 .d_write = physwrite,
109 .d_strategy = vnstrategy,
115 int sc_flags; /* flags */
116 u_int64_t sc_size; /* size of vn, sc_secsize scale */
117 int sc_secsize; /* sector size */
118 struct diskslices *sc_slices; /* XXX fields from struct disk */
119 struct disk_info sc_info; /* XXX fields from struct disk */
120 struct vnode *sc_vp; /* vnode if not NULL */
121 vm_object_t sc_object; /* backing object if not NULL */
122 struct ucred *sc_cred; /* credentials */
123 int sc_maxactive; /* max # of active requests */
124 struct buf sc_tab; /* transfer queue */
125 u_long sc_options; /* options */
126 cdev_t sc_devlist; /* devices that refer to this unit */
127 SLIST_ENTRY(vn_softc) sc_list;
130 static SLIST_HEAD(, vn_softc) vn_list;
133 #define VNF_INITED 0x01
134 #define VNF_READONLY 0x02
136 static u_long vn_options;
138 #define IFOPT(vn,opt) if (((vn)->sc_options|vn_options) & (opt))
139 #define TESTOPT(vn,opt) (((vn)->sc_options|vn_options) & (opt))
141 static int vnsetcred (struct vn_softc *vn, struct ucred *cred);
142 static void vnclear (struct vn_softc *vn);
143 static int vn_modevent (module_t, int, void *);
144 static int vniocattach_file (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
145 static int vniocattach_swap (struct vn_softc *, struct vn_ioctl *, cdev_t dev, int flag, struct ucred *cred);
148 vnclose(struct dev_close_args *ap)
150 cdev_t dev = ap->a_head.a_dev;
151 struct vn_softc *vn = dev->si_drv1;
154 if (vn->sc_slices != NULL)
155 dsclose(dev, ap->a_devtype, vn->sc_slices);
160 * Called only when si_drv1 is NULL. Locate the associated vn node and
161 * attach the device to it.
163 static struct vn_softc *
170 SLIST_FOREACH(vn, &vn_list, sc_list) {
171 if (vn->sc_unit == unit) {
173 dev->si_drv2 = vn->sc_devlist;
174 vn->sc_devlist = dev;
180 vn = kmalloc(sizeof *vn, M_DEVBUF, M_WAITOK | M_ZERO);
183 vn->sc_devlist = make_dev(&vn_ops, 0, UID_ROOT,
184 GID_OPERATOR, 0640, "vn%d", unit);
185 if (vn->sc_devlist->si_drv1 == NULL) {
186 reference_dev(vn->sc_devlist);
187 vn->sc_devlist->si_drv1 = vn;
188 vn->sc_devlist->si_drv2 = NULL;
190 if (vn->sc_devlist != dev) {
192 dev->si_drv2 = vn->sc_devlist;
193 vn->sc_devlist = dev;
196 SLIST_INSERT_HEAD(&vn_list, vn, sc_list);
202 vnopen(struct dev_open_args *ap)
204 cdev_t dev = ap->a_head.a_dev;
206 struct disk_info *info;
209 * Locate preexisting device
212 if ((vn = dev->si_drv1) == NULL)
216 * Update si_bsize fields for device. This data will be overriden by
217 * the slice/parition code for vn accesses through partitions, and
218 * used directly if you open the 'whole disk' device.
220 * si_bsize_best must be reinitialized in case VN has been
221 * reconfigured, plus make it at least VN_BSIZE_BEST for efficiency.
223 dev->si_bsize_phys = vn->sc_secsize;
224 dev->si_bsize_best = vn->sc_secsize;
225 if (dev->si_bsize_best < VN_BSIZE_BEST)
226 dev->si_bsize_best = VN_BSIZE_BEST;
228 if ((ap->a_oflags & FWRITE) && (vn->sc_flags & VNF_READONLY))
232 kprintf("vnopen(%s, 0x%x, 0x%x)\n",
233 devtoname(dev), ap->a_oflags, ap->a_devtype);
239 IFOPT(vn, VN_LABELS) {
240 if (vn->sc_flags & VNF_INITED) {
242 bzero(info, sizeof(*info));
243 info->d_media_blksize = vn->sc_secsize;
244 info->d_media_blocks = vn->sc_size;
246 * reserve mbr sector for backwards compatibility
247 * when no slices exist.
249 info->d_dsflags = DSO_COMPATMBR;
251 info->d_secpertrack = 32;
252 info->d_nheads = 64 / (vn->sc_secsize / DEV_BSIZE);
253 info->d_secpercyl = info->d_secpertrack *
255 info->d_ncylinders = vn->sc_size / info->d_secpercyl;
257 return (dsopen(dev, ap->a_devtype, 0,
258 &vn->sc_slices, info));
260 if (dkslice(dev) != WHOLE_DISK_SLICE ||
261 dkpart(dev) != WHOLE_SLICE_PART ||
262 ap->a_devtype != S_IFCHR) {
272 * Run strategy routine for VN device. We use VOP_READ/VOP_WRITE calls
273 * for vnode-backed vn's, and the new vm_pager_strategy() call for
274 * vm_object-backed vn's.
276 * Currently B_ASYNC is only partially handled - for OBJT_SWAP I/O only.
279 vnstrategy(struct dev_strategy_args *ap)
281 cdev_t dev = ap->a_head.a_dev;
282 struct bio *bio = ap->a_bio;
290 if ((vn = dev->si_drv1) == NULL)
296 kprintf("vnstrategy(%p): unit %d\n", bp, unit);
298 if ((vn->sc_flags & VNF_INITED) == 0) {
300 bp->b_flags |= B_ERROR;
305 bp->b_resid = bp->b_bcount;
307 IFOPT(vn, VN_LABELS) {
309 * The vnode device is using disk/slice label support.
311 * The dscheck() function is called for validating the
312 * slices that exist ON the vnode device itself, and
313 * translate the "slice-relative" block number, again.
314 * dscheck() will call biodone() and return NULL if
315 * we are at EOF or beyond the device size.
317 if (vn->sc_slices == NULL) {
319 } else if ((nbio = dscheck(dev, bio, vn->sc_slices)) == NULL) {
323 int64_t pbn; /* in sc_secsize chunks */
324 long sz; /* in sc_secsize chunks */
327 * Check for required alignment. Transfers must be a valid
328 * multiple of the sector size.
330 if (bp->b_bcount % vn->sc_secsize != 0 ||
331 bio->bio_offset % vn->sc_secsize != 0) {
335 pbn = bio->bio_offset / vn->sc_secsize;
336 sz = howmany(bp->b_bcount, vn->sc_secsize);
339 * Check for an illegal pbn or EOF truncation
343 if (pbn + sz > vn->sc_size) {
344 if (pbn > vn->sc_size || (bp->b_flags & B_BNOCLIP))
346 if (pbn == vn->sc_size) {
347 bp->b_resid = bp->b_bcount;
348 bp->b_flags |= B_INVAL;
351 bp->b_bcount = (vn->sc_size - pbn) * vn->sc_secsize;
353 nbio = push_bio(bio);
354 nbio->bio_offset = pbn * vn->sc_secsize;
358 * Use the translated nbio from this point on
360 if (vn->sc_vp && bp->b_cmd == BUF_CMD_FREEBLKS) {
362 * Freeblks is not handled for vnode-backed elements yet.
365 /* operation complete */
366 } else if (vn->sc_vp) {
370 * If an error occurs, we set B_ERROR but we do not set
371 * B_INVAL because (for a write anyway), the buffer is
377 bzero(&auio, sizeof(auio));
379 aiov.iov_base = bp->b_data;
380 aiov.iov_len = bp->b_bcount;
381 auio.uio_iov = &aiov;
383 auio.uio_offset = nbio->bio_offset;
384 auio.uio_segflg = UIO_SYSSPACE;
385 if (bp->b_cmd == BUF_CMD_READ)
386 auio.uio_rw = UIO_READ;
388 auio.uio_rw = UIO_WRITE;
389 auio.uio_resid = bp->b_bcount;
390 auio.uio_td = curthread;
391 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
392 if (bp->b_cmd == BUF_CMD_READ)
393 error = VOP_READ(vn->sc_vp, &auio, IO_DIRECT, vn->sc_cred);
395 error = VOP_WRITE(vn->sc_vp, &auio, IO_NOWDRAIN, vn->sc_cred);
396 vn_unlock(vn->sc_vp);
397 bp->b_resid = auio.uio_resid;
400 bp->b_flags |= B_ERROR;
402 /* operation complete */
403 } else if (vn->sc_object) {
405 * OBJT_SWAP I/O (handles read, write, freebuf)
407 * We have nothing to do if freeing blocks on a reserved
408 * swap area, othrewise execute the op.
410 if (bp->b_cmd == BUF_CMD_FREEBLKS && TESTOPT(vn, VN_RESERVE)) {
412 /* operation complete */
414 vm_pager_strategy(vn->sc_object, nbio);
419 bp->b_resid = bp->b_bcount;
420 bp->b_flags |= B_ERROR | B_INVAL;
421 bp->b_error = EINVAL;
422 /* operation complete */
428 * Shortcuts / check failures on the original bio (not nbio).
431 bp->b_error = EINVAL;
432 bp->b_flags |= B_ERROR | B_INVAL;
440 vnioctl(struct dev_ioctl_args *ap)
442 cdev_t dev = ap->a_head.a_dev;
444 struct vn_ioctl *vio;
449 IFOPT(vn,VN_FOLLOW) {
450 kprintf("vnioctl(%s, 0x%lx, %p, 0x%x): unit %d\n",
451 devtoname(dev), ap->a_cmd, ap->a_data, ap->a_fflag,
465 IFOPT(vn,VN_LABELS) {
466 if (vn->sc_slices != NULL) {
467 error = dsioctl(dev, ap->a_cmd, ap->a_data,
469 &vn->sc_slices, &vn->sc_info);
470 if (error != ENOIOCTL)
473 if (dkslice(dev) != WHOLE_DISK_SLICE ||
474 dkpart(dev) != WHOLE_SLICE_PART)
480 error = suser_cred(ap->a_cred, 0);
484 vio = (struct vn_ioctl *)ap->a_data;
485 f = (u_long*)ap->a_data;
489 if (vn->sc_flags & VNF_INITED)
492 if (vio->vn_file == NULL)
493 error = vniocattach_swap(vn, vio, dev, ap->a_fflag, ap->a_cred);
495 error = vniocattach_file(vn, vio, dev, ap->a_fflag, ap->a_cred);
499 if ((vn->sc_flags & VNF_INITED) == 0)
502 * XXX handle i/o in progress. Return EBUSY, or wait, or
504 * XXX handle multiple opens of the device. Return EBUSY,
505 * or revoke the fd's.
506 * How are these problems handled for removable and failing
507 * hardware devices? (Hint: They are not)
511 kprintf("vnioctl: CLRed\n");
525 vn->sc_options |= *f;
530 vn->sc_options &= ~(*f);
544 * Attach a file to a VN partition. Return the size in the vn_size
549 vniocattach_file(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
550 int flag, struct ucred *cred)
553 struct nlookupdata nd;
557 flags = FREAD|FWRITE;
558 error = nlookup_init(&nd, vio->vn_file,
559 UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
562 if ((error = vn_open(&nd, NULL, flags, 0)) != 0) {
563 if (error != EACCES && error != EPERM && error != EROFS)
567 error = nlookup_init(&nd, vio->vn_file, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
570 if ((error = vn_open(&nd, NULL, flags, 0)) != 0)
574 if (vp->v_type != VREG ||
575 (error = VOP_GETATTR(vp, &vattr))) {
581 vn->sc_secsize = DEV_BSIZE;
583 nd.nl_open_vp = NULL;
586 * If the size is specified, override the file attributes. Note that
587 * the vn_size argument is in PAGE_SIZE sized blocks.
590 vn->sc_size = vio->vn_size * PAGE_SIZE / vn->sc_secsize;
592 vn->sc_size = vattr.va_size / vn->sc_secsize;
593 error = vnsetcred(vn, cred);
599 vn->sc_flags |= VNF_INITED;
601 vn->sc_flags |= VNF_READONLY;
602 IFOPT(vn, VN_LABELS) {
604 * Reopen so that `ds' knows which devices are open.
605 * If this is the first VNIOCSET, then we've
606 * guaranteed that the device is the cdev and that
607 * no other slices or labels are open. Otherwise,
608 * we rely on VNIOCCLR not being abused.
610 error = dev_dopen(dev, flag, S_IFCHR, cred);
615 kprintf("vnioctl: SET vp %p size %llx blks\n",
616 vn->sc_vp, vn->sc_size);
625 * Attach swap backing store to a VN partition of the size specified
630 vniocattach_swap(struct vn_softc *vn, struct vn_ioctl *vio, cdev_t dev,
631 int flag, struct ucred *cred)
636 * Range check. Disallow negative sizes or any size less then the
637 * size of a page. Then round to a page.
640 if (vio->vn_size <= 0)
644 * Allocate an OBJT_SWAP object.
646 * sc_secsize is PAGE_SIZE'd
648 * vio->vn_size is in PAGE_SIZE'd chunks.
649 * sc_size must be in PAGE_SIZE'd chunks.
650 * Note the truncation.
653 vn->sc_secsize = PAGE_SIZE;
654 vn->sc_size = vio->vn_size;
655 vn->sc_object = vm_pager_allocate(OBJT_SWAP, NULL,
656 vn->sc_secsize * (off_t)vio->vn_size,
658 IFOPT(vn, VN_RESERVE) {
659 if (swap_pager_reserve(vn->sc_object, 0, vn->sc_size) < 0) {
660 vm_pager_deallocate(vn->sc_object);
661 vn->sc_object = NULL;
665 vn->sc_flags |= VNF_INITED;
667 error = vnsetcred(vn, cred);
669 IFOPT(vn, VN_LABELS) {
671 * Reopen so that `ds' knows which devices are open.
672 * If this is the first VNIOCSET, then we've
673 * guaranteed that the device is the cdev and that
674 * no other slices or labels are open. Otherwise,
675 * we rely on VNIOCCLR not being abused.
677 error = dev_dopen(dev, flag, S_IFCHR, cred);
681 IFOPT(vn, VN_FOLLOW) {
682 kprintf("vnioctl: SET vp %p size %llx\n",
683 vn->sc_vp, vn->sc_size);
692 * Duplicate the current processes' credentials. Since we are called only
693 * as the result of a SET ioctl and only root can do that, any future access
694 * to this "disk" is essentially as root. Note that credentials may change
695 * if some other uid can write directly to the mapped file (NFS).
698 vnsetcred(struct vn_softc *vn, struct ucred *cred)
704 * Set credits in our softc
709 vn->sc_cred = crdup(cred);
712 * Horrible kludge to establish credentials for NFS XXX.
719 tmpbuf = kmalloc(vn->sc_secsize, M_TEMP, M_WAITOK);
720 bzero(&auio, sizeof(auio));
722 aiov.iov_base = tmpbuf;
723 aiov.iov_len = vn->sc_secsize;
724 auio.uio_iov = &aiov;
727 auio.uio_rw = UIO_READ;
728 auio.uio_segflg = UIO_SYSSPACE;
729 auio.uio_resid = aiov.iov_len;
730 vn_lock(vn->sc_vp, LK_EXCLUSIVE | LK_RETRY);
731 error = VOP_READ(vn->sc_vp, &auio, 0, vn->sc_cred);
732 vn_unlock(vn->sc_vp);
733 kfree(tmpbuf, M_TEMP);
739 vnclear(struct vn_softc *vn)
742 kprintf("vnclear(%p): vp=%p\n", vn, vn->sc_vp);
743 if (vn->sc_slices != NULL)
744 dsgone(&vn->sc_slices);
745 vn->sc_flags &= ~VNF_INITED;
746 if (vn->sc_vp != NULL) {
748 (vn->sc_flags & VNF_READONLY) ? FREAD : (FREAD|FWRITE));
751 vn->sc_flags &= ~VNF_READONLY;
756 if (vn->sc_object != NULL) {
757 vm_pager_deallocate(vn->sc_object);
758 vn->sc_object = NULL;
764 vnsize(struct dev_psize_args *ap)
766 cdev_t dev = ap->a_head.a_dev;
772 if ((vn->sc_flags & VNF_INITED) == 0)
774 ap->a_result = (int64_t)vn->sc_size;
779 vn_modevent(module_t mod, int type, void *data)
786 dev_ops_add(&vn_ops, 0, 0);
791 while ((vn = SLIST_FIRST(&vn_list)) != NULL) {
792 SLIST_REMOVE_HEAD(&vn_list, sc_list);
793 if (vn->sc_flags & VNF_INITED)
795 /* Cleanup all cdev_t's that refer to this unit */
796 while ((dev = vn->sc_devlist) != NULL) {
797 vn->sc_devlist = dev->si_drv2;
798 dev->si_drv1 = dev->si_drv2 = NULL;
803 dev_ops_remove(&vn_ops, 0, 0);
811 DEV_MODULE(vn, vn_modevent, 0);